LuxCoreRender Forums

blibli wrote: ↑Sat Jan 20, 2018 6:01 pm Would it be possible to build both a 3.5 and 3.6 version? I'm pretty sure 2.79a will use latest svn with all deps updated because of the vulnerabilities fuzz/scandal.

It would help a lot to know if 2.79a is going to use Python 3.5 or 3.6. Isn't there someone in Blender Foundation that can answer the question

Normally that should not be an issue as long as no newer py exclusive functions are used ( which are not afaik ).
On Linux and macOS i made it so the module pulls the python from Blender itself.
Dunno if thats possible with Windows, someone should take a look.

The trick is to use dynamic_lookup on macOS or same default link behaviour on Linux and not even link pythonlib.
Maybe Win need a flag set for such like macOS does. ( equivalent to: -undefined dynamic_lookup )
More to readup here: https://msdn.microsoft.com/en-us/librar ... s.85).aspx

Jens

Dade wrote: ↑Sat Jan 20, 2018 6:09 pm

blibli wrote: ↑Sat Jan 20, 2018 6:01 pm Would it be possible to build both a 3.5 and 3.6 version? I'm pretty sure 2.79a will use latest svn with all deps updated because of the vulnerabilities fuzz/scandal.

It would help a lot to know if 2.79a is going to use Python 3.5 or 3.6. Isn't there someone in Blender Foundation that can answer the question

So after asking on IRC, they will decide tomorrow. The rule is to not update libs, but they have to balance it with all the bad publicity they got I guess.

blibli wrote: ↑Sun Jan 21, 2018 12:12 pm So after asking on IRC, they will decide tomorrow. The rule is to not update libs,

It is quite the ABC of software release.

blibli wrote: ↑Sun Jan 21, 2018 12:12 pm but they have to balance it with all the bad publicity they got I guess.

Because that security-thing I have read around

jensverwiebe wrote: ↑Sat Jan 20, 2018 8:15 pm On Linux and macOS i made it so the module pulls the python from Blender itself.
Dunno if thats possible with Windows, someone should take a look.

How is that done in Linux and MacOS? Is there a call to a function similar to what you've linked for microsoft?

jensverwiebe wrote: ↑Sat Jan 20, 2018 8:15 pm The trick is to use dynamic_lookup on macOS or same default link behaviour on Linux and not even link pythonlib.
Maybe Win need a flag set for such like macOS does. ( equivalent to: -undefined dynamic_lookup )
More to readup here: https://msdn.microsoft.com/en-us/librar ... s.85).aspx

My understanding from the cmake files and pyluxcore code is that python lib is never referenced and linked directly, but only through boost.
Is that correct? Then also boost should not be required to link statically to python?

acasta69 wrote: ↑Sun Jan 21, 2018 2:51 pm

jensverwiebe wrote: ↑Sat Jan 20, 2018 8:15 pm On Linux and macOS i made it so the module pulls the python from Blender itself.
Dunno if thats possible with Windows, someone should take a look.

How is that done in Linux and MacOS? Is there a call to a function similar to what you've linked for microsoft?

jensverwiebe wrote: ↑Sat Jan 20, 2018 8:15 pm The trick is to use dynamic_lookup on macOS or same default link behaviour on Linux and not even link pythonlib.
Maybe Win need a flag set for such like macOS does. ( equivalent to: -undefined dynamic_lookup )
More to readup here: https://msdn.microsoft.com/en-us/librar ... s.85).aspx

My understanding from the cmake files and pyluxcore code is that python lib is never referenced and linked directly, but only through boost.
Is that correct? Then also boost should not be required to link statically to python?

Its cause a module "talks" 2-way with a host. But i really dunno anything about win in this regards.
I doubt you can leave out boost, cause luxcore for sure uses more/other symbols than linked into blender.
Blender relases come with static boost, so only used symbols are to be found ( which explains python works this way )

Jens

Dade wrote: ↑Sun Jan 21, 2018 12:44 pm

blibli wrote: ↑Sun Jan 21, 2018 12:12 pm So after asking on IRC, they will decide tomorrow. The rule is to not update libs,

It is quite the ABC of software release.

Of course, they know it and they are the first wanting 2.79a to be stable to concentrate on 2.8x but...

Dade wrote: ↑Sun Jan 21, 2018 12:44 pm

blibli wrote: ↑Sun Jan 21, 2018 12:12 pm but they have to balance it with all the bad publicity they got I guess.

Because that security-thing I have read around

Yes

This gonna be by far my longest awaiting from a blender official release. i feel so compressed while waiting so ...........................to put my hand on the Engine and shoot a real Q3 track time

Sharlybg wrote: ↑Mon Jan 22, 2018 6:46 pm This gonna be by far my longest awaiting from a blender official release. i feel so compressed while waiting so ...........................to put my hand on the Engine and shoot a real Q3 track time

Why waiting ? Grab an actual testbuild, else we don't get rid of hidden bugs.
Just avoid using such builds as production sw.

Edit: just saw this time we had no testbuilds , but they should show up asap. Today awaiting an "AHOI".
Check here: http://download.blender.org/release/Blender2.79/

Jens

Edit: just saw this time we had no testbuilds , but they should show up asap. Today awaiting an "AHOI".
Check here: http://download.blender.org/release/Blender2.79/

Jens

I hope so. Also hope that when testbuilds will be there blender will not refuse to install blendluxcore. I've try with this builds
( https://builder.blender.org/download/ )
and it return error .